Prediction of Solar Radiation Using Meteorological Data

DEMİRTAŞ M., Yesilbudak M., Sagiroglu Ş., Colak I.

1st International Conference on Renewable Energy Research and Applications (ICRERA), Nagasaki, Japonya, 11 - 14 Kasım 2012 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası:
  • Doi Numarası: 10.1109/icrera.2012.6477329
  • Basıldığı Şehir: Nagasaki
  • Basıldığı Ülke: Japonya
  • Anahtar Kelimeler: Meteorological data, solar radiation, instance-based learning, short-term prediction
  • Gazi Üniversitesi Adresli: Evet

Özet

Solar radiation prediction has a great importance in electricity generation from solar energy and helps to size photovoltaic power systems. Therefore, the solar radiation parameter was predicted at 10-min intervals in this study. Outside temperature, outside humidity and barometric pressure parameters were used as meteorological input variables by the developed k-nearest neighbor (k-NN) classifier. On the one hand, it is mined that solar radiation prediction was affected by the number of nearest neighbors, the dimension of input parameters and the type of distance metrics. On the other hand, it is shown that the k-NN classifier which uses Euclidean distance metric for k=4 in 3-dimensional input space outperformed the other models in terms of the prediction accuracy. Adversely, the k-NN classifier which only uses barometric pressure input provided the weakest prediction performance for k=15 in Euclidean distance metric.